polyfit error: Subscripted assignment dimension mismatch.

I am trying to fit a relationship between temperature and salinity values for different depth bins. At the moment polyfit is not working. If I use degree=1, the error I get is:

Subscripted assignment dimension mismatch. Error in temp_sal_plot_K_edits_v2 (line 88) p(i,:)=polyfit(T,S,1); % Find polynomial relationship for this (i) depth bin

Plus I always get a warning message of this:

Warning: Polynomial is badly conditioned. Add points with distinct X values, reduce the degree of the polynomial, or try centering and scaling as described in HELP POLYFIT. > In polyfit at 75 In temp_sal_plot_K_edits_v2 at 40

If I change the degree to 3, polyfit runs, however the fitted polynomial is very inaccurate. Could someone please let me know where I am going wrong? Here is the code:

for i=300:310 % Loop over m steps in depth
    T=[];
    S=[];
    for j=1:nx % Loop over degC steps in tempearure 
        if vS(i,j) > 0
            T=[T qX(i,j)]; % Temperature = temperature value from query points
            S=[S vS(i,j)]; % Salinty = salinity value from interpolated grid
        end

      end

      p(i,:)=polyfit(T,S,1); % Find polynomial relationship for this (i) depth bin
      fS(i,:)=polyval(p(i,:),qX(i,:)); % Evaluate fS points (Salinity points) for each query point (qX = temperature points) using the derived polynomial expression for this depth bin.
      figure;
      plot(S,T,'o',qX(i,:),fS(i,:));
      xlabel('Salinity, ppt');
      ylabel('Temperature, degC');

end

Answer 1

[Matt J]

The error message isn't saying that polyfit is at fault. It's saying that p(i,:) is not the same size as polyfit(T,S,1) in this line

   p(i,:)=polyfit(T,S,1);

Likely, you are are trying to stuff the 1x2 output of polyfit into the rows of a pre-existing M x N matrix p where N is not equal to 2. In fact, my guess is that N=4 since you say you got it to run with cubic polynomials.

As for the warning, there's not much we can say without seeing the contents of the T,S that are triggering it. In all likelihood, your T,S data is just bad and isn't well-fit by a unique polynomial. However, you could try doing as the warning instructs and improve the scaling/centering of the data, or add more points.